Ok, this one I'm applying. I agree that it leaves MS_ASYNC not updating
the file until the next sync actually happens, but I can't really bring
myself to care at least for an imminent 2.6.24 thing. The file times are
actually "correct" in the sense that they will now match when the IO is
done, and my man-page says that MS_ASYNC "schedules the io to be done".
And I think this is better than we have now, and I don't think this part
is somethign that anybody really disagrees with.
We can (and should) keep the MS_ASYNC issue open.
Linus
--